Information presentation over the Internet is changing dramatically. New time-varying multimedia content is now being brought to the Internet, and in particular to the World Wide Web (the web), in addition to textual HTML pages and still graphics. Here, time-varying multimedia content refers to sound, video, animated graphics, or any other medium that evolves as a function of elapsed time, alone or in combination. In many situations, instant delivery and presentation of such multimedia content, on demand, is desired.
“On-demand” is a term for a wide set of technologies that enable individuals to select multimedia content from a central server for instant delivery and presentation on a client (computer or television). For example, video-on-demand can be used for entertainment (ordering movies transmitted digitally), education (viewing training videos) and browsing (viewing informative audiovisual material on a web page) to name a few examples.
Users are generally connected to the Internet by a communications link of limited bandwidth, such as a 56 kilo bits per second (Kbps) modem or an integrated services digital network (ISDN) connection. Even corporate users are usually limited to a fraction of the 1.544 mega bits per second (Mbps) T-1 carrier rates. This bandwidth limitation provides a challenge to on-demand systems: it may be impossible to transmit a large amount of image or video data over a limited bandwidth in the short amount of time required for “instant delivery and presentation.” Downloading a large image or video may take hours before presentation can begin. As a consequence, special techniques have been developed for on-demand processing of large images and video.
A technique for providing large images on demand over a communications link with limited bandwidth is progressive image transmission. In progressive image transmission, each image is encoded, or compressed, in layers, like an onion. The first (core) layer, or base layer, represents a low-resolution version of the image. Successive layers represent successively higher resolution versions of the image. The server transmits the layers in order, starting from the base layer. The client receives the base layer, and instantly presents to the user a low-resolution version of the image. The client presents higher resolution versions of the image as the successive layers are received. Progressive image transmission enables the user to interact with the server instantly, with low delay, or low latency. For example, progressive image transmission enables a user to browse through a large database of images, quickly aborting the transmission of the unwanted images before they are completely downloaded to the client.
Similarly, streaming is a technique that provides time-varying content, such as video and audio, on demand over a communications link with limited bandwidth. In streaming, audiovisual data is packetized, delivered over a network, and played as the packets are being received at the receiving end, as opposed to being played only after all packets have been downloaded. Streaming technologies are becoming increasingly important with the growth of the Internet because most users do not have fast enough access to download large multimedia files quickly. With streaming, the client browser or application can start displaying the data before the entire file has been transmitted.
In a video on-demand delivery system that uses streaming, the audiovisual data is often compressed and stored on a disk on a media server for later transmission to a client system. For streaming to work, the client side receiving the data must be able to collect the data and send it as a steady stream to a decoder or an application that is processing the data and converting it to sound or pictures. If the client receives the data more quickly than required, it needs to save the excess data in a buffer. Conversely, if the client receives the data more slowly than required, it needs to play out some of the data from the buffer. Storing part of a multimedia file in this manner before playing the file is referred to as buffering. Buffering can provide smooth playback even if the client temporarily receives the data more quickly or more slowly than required for real-time playback.
There are two reasons that a client can temporarily receive data more quickly or more slowly than required for real-time playback. First, in a variable-rate transmission system such as a packet network, the data arrives at uneven rates. Not only does packetized data inherently arrive in bursts, but even packets of data that are transmitted from the sender at an even rate may not arrive at the receiver at an even rate. This is due to the fact that individual packets may follow different routes, and the delay through any individual router may vary depending on the amount of traffic waiting to go through the router. The variability in the rate at which data is transmitted through a network is called network jitter.
A second reason that a client can temporarily receive data more quickly or more slowly than required for real-time playback is that the media content is encoded to variable bit rate. For example, high-motion scenes in a video may be encoded with more bits than low-motion scenes. When the encoded video is transmitted with a relatively constant bit rate, then the high-motion frames arrive at a slower rate than the low-motion frames. For both these reasons (variable-rate source encoding and variable-rate transmission channels), buffering is required at the client to allow a smooth presentation.
Unfortunately, buffering implies delay, or latency. Start-up delay refers to the latency the user experiences after he signals the server to start transmitting data from the beginning of the content (such as when a pointer to the content is selected by the user) before the data can be decoded by the client system and presented to the user. Seek delay refers to the latency the user experiences after he signals to the server to start transmitting data from an arbitrary place in the middle of the content (such as when a seek bar is dragged to a particular point in time) before the data can be decoded and presented. Both start-up and seek delays occur because even after the client begins to receive new data, it must wait until its buffer is sufficiently full to begin playing out of the buffer. It does this in order to guard against future buffer underflow due to network jitter and variable-bit rate compression. For typical audiovisual coding on the Internet, start-up and seek delays between two and ten seconds are common.
Large start-up and seek delays are particularly annoying when the user is trying to browse through a large amount of audiovisual content trying to find a particular video or a particular location in a video. As in the image browsing scenario using progressive transmission, most of the time the user will want to abort the transmission long before all the data are downloaded and presented. In such a scenario, delays of two to ten seconds between aborts seem intolerable. What is needed is a method for reducing the start-up and seek delays for such “on demand” interactive multimedia applications.